Safety aspects are of fundamental importance while developing new automobiles. Aims of passive safety are the reduction of the physical injury risk and the level of injury for every car occupant. For reaching these aims, a very early knowledge of the further sequence of accident events is important and necessary. Thus different restraint systems can be used to meet the specification. Furthermore an individually adjusted activation of the restraint systems dependent on the accident scenario is possible. To predict the further sequence of accident events, it is important to obtain knowledge about the impact speed, the barrier position and the impact angle. This paper introduces an approach to determine the three accident parameters mentioned above by using signals of common sensors of modern automobiles only. For an early sequence of accident events prediction just the first 10 to 20 ms after the initial contact with the barrier of different acceleration signals are used. Subsequent to that the acceleration signals are wavelet transformed by using a complex Gaussian wavelet of 4th order. In a following step the results of the wavelet transformation are strongly discretized in time and frequency domain and are shown in figure 1. The calculated wavelet coefficients of the different acceleration sensors are conduced as input parameters of an artificial neural network. For a good determination of every accident parameter, separate neural networks are necessary. This paper presents the results of a pole barrier study with a Euro NCAP pole as the obstacle in front of the automobile. During the study about 1250 FE full vehicle simulations were done in which impact velocities, impact angles and barrier positions were varied. For the subsequent determination of the accident parameters, different sensors were defined inside the FE-model, e.g. at the longitudinal chassis beams and at the chassis tunnel. Furthermore, the yaw rate and the resultant yaw angle are taken into account to improve the results. The procedure can be used to classify different accident scenarios and their accident parameters, e.g. the standard deviation of the impact velocity adds up to approximately 1.4 km/h. In addition an accurate determination of impact angle and barrier position is possible.
Categorisation of accident scenarios by using wavelet transformed signals of FE-crash simulations and artificial neural networks
Kategorisierung von Unfallszenarien durch Wavelet-transformierte Signale von FE-Crash-Simulationen und künstlichen neuralen Netzwerken
2011
12 Seiten, 10 Bilder, 8 Quellen
(nicht paginiert)
Conference paper
Storage medium
English
Automobilbau , Sicherheitsfaktor , passive Sicherheit , Kategorisierung , Spezifikation (Normung) , Aufprallgeschwindigkeit , Verkehrsunfallursache , Aufprallverhalten , Beschleunigungssensor , Signalverarbeitung , Wavelet-Transformation , Zeitbereichanalyse , Frequenzbereichsanalyse , Parameteridentifikation
Open-World Learning for Traffic Scenarios Categorisation
IEEE | 2023
|Simulation of real pre-crash accident scenarios using German In-Depth Accident Study ( GIDAS )
Automotive engineering | 2009
|Simulation of real pre crash accident scenarios using German in-depth accident study (GIDAS)
ELBA - Federal Highway Research Institute (BASt) | 2012
|Generic Crash Pulses Representing Future Accident Scenarios of Highly Automated Vehicles
SAE Technical Papers | 2022
|Crash phase accident reconstruction
Automotive engineering | 1986
|